Wavelet compression with set partitioning for low bandwidth telemetry from AUVs

Abstract

Autonomous underwater vehicles (AUVs) typically communicate with scientists on the surface over an unreliable wireless channel. The challenges of underwater acoustic communication result in very low data throughput. While there are several examples of scientific data, even imagery, being successfully transmitted over high rate acoustic links, channel coding methods with high rates of error-correction are often employed that limit data throughput to tens or a few hundred bits per second. Little research exists into appropriate methods for image and data compression for acoustic links at these very low rates. We recently have experienced great success using compression techniques based upon the Set Partitioning in Hierarchical Trees (SPIHT) embedded coding method, and feel they are particularly suited to underwater data in a number of ways. In particular, SPIHT provides a fully embedded coding method; truncating the encoded bitstream at any point produces the optimal encoding for that data length. This allows fine-resolution imagery to build on previously transmitted low-resolution thumbnails. For time-series data, we have developed a method for quantizing data to emphasize more important sections, such as the most recently collected data. In this paper we describe how these methods can be applied to compress scalar environmental data and imagery for communication over acoustic links. We also the present initial results of sea trials performed near Rota in the Commonwealth of Northern Marianas Islands, during which images were captured, compressed and transmitted in-situ.